Economic attention allocation
Each Atom has an Attention Value attached to it. The process of updating these values is carried out according to nonlinear dynamical equations that are derived based on "artificial economics," utilizing two separate "currencies," one for Short Term Importance (STI) and one for Long Term Importance (LTI).
One aspect of these equations is a form of Hebbian Learning: Atoms called HebbianLinks record which Atoms were habitually used together in the past, and when it occurred that Atom A's utilization appeared to play a role in causing Atom B's utilization. Then, these HebbianLinks are used to guide the flow of currency between Atoms: B gives A some money if B thinks that this money will help A to get used, and that this utilization will help B to get used.
Very roughly speaking, these dynamical equations play a similar role to that played by activation-spreading in Neural Network AI systems.
This is an important process because
- most MindAgents select which Atoms to act upon via a choice function that is biased based on Short Term Importance
- the Forgetting MindAgent utilizes "low Long Term Importance as a criterion for removing Atoms from the AtomTable, thus eliminating them from Long Term Memory (and either deleting them permanently or pushing them to Ultra Long Term Memory).